Global Climate Models (GCMs) are indispensable for climate impact assessment, yet their direct application at regional scales is limited by systematic biases and weak representation of temporal persistence. This study develops a reproducible and transferable framework to evaluate and enhance the reliability of CMIP6 GCM precipitation for hydroclimatic applications in Bangladesh, using long-term monthly observations from 35 meteorological stations. Two contrasting models, ACCESS-ESM1-5 and CanESM5, are assessed against observed mean, variability, temporal persistence, and extremes. Systematic biases are quantified and corrected using empirical Quantile Mapping (eQM) and Nested Bias Correction (NBC), with explicit treatment of lag-1 autocorrelation and interannual variability—features commonly overlooked in conventional correction approaches. Results indicate that ACCESS-ESM1-5 better reproduces variability and extremes despite spatially heterogeneous biases, whereas CanESM5 exhibits a persistent dry bias and underestimates high-intensity rainfall. Comparative analysis shows that NBC outperforms eQM by jointly correcting mean, variance, and persistence, yielding hydro-climatically consistent precipitation sequences suitable for impact modeling. Application under SSP2-4.5 and SSP5-8.5 scenarios demonstrates that persistence-aware correction substantially modifies projected precipitation signals. The proposed framework transforms raw GCM outputs into actionable regional climate information for hydrological design, flood risk assessment, and climate adaptation planning in climate-vulnerable regions.
Nury et al. (2026) studied this question.